If we want to know what news organisations will look like under the second Trump administration in America, well, we are beginning to get an idea. At the Dealbook conference in New York last week, Jeff Bezos, the owner of the Washington Post and multi-billionaire founder of Amazon, gave a very favourable assessment of Donald Trump’s upcoming second term. “I am very hopeful … he seems to have a lot of energy around reducing regulation,” beamed Bezos. It was surprising, then, that the Washington Post did not endorse Trump in its pre-election editorial. Instead the writers crafted an endorsement of Democratic candidate Kamala Harris which Bezos killed, in his first act of blatant editorial interference since he bought the title in 2013. In a Post article Bezos rationalised his decision as being an attempt to restore trust in the press with what he called “independence” – a type of independence which clearly does not extend to damping down public flattery. The headline, run with Bezos’s approval, was ‘The Hard Truth: Why Americans Don’t Trust the News Media’. Bezos invoked the past of the American press, its historic non-endorsement of candidates and a return to some past notion of “objectivity” as being a fix for his news organisation’s woes.
Patrick Soon-Shiong, the billionaire owner of the Los Angeles Times, had already emulated Bezos by stopping the LA Times editorial board from running a presidential endorsement. Soon-Shiong, appearing on the Republican commentator Scott Jennings’ radio show, said that he had been working on an AI-powered “bias meter” which will appear on LA Times articles from January. The motivation was that the medical tech billionaire said he had begun to see his own news title as “an echo chamber and not a trusted source”.
Bezos and Soon-Shiong are technology company billionaires, who have built enormous fortunes on activities where the public trust is low. The least trustworthy among them are even ranked below the press, according to a 2021 confidence survey which also showed Amazon’s trust levels dropping more dramatically than any other institution surveyed. But the point is not the metric or the calibration. The point is to reinforce the owners’ own views at a time when aligning themselves more with the right is far more politically expedient for their more profitably untrustworthy companies.
Turning to AI to manufacture “objectivity” is revealing in itself. Princeton psychology professor Molly Crockett has studied and written extensively about how technological systems and human behaviour interact to impact society. She says, “the AI bias meter is such a great example of the ‘Oracle of AI’ . It is not grounded in science and we don’t have the ability for AI to detect bias at article level. It does not exist”. Earlier this year Crockett, along with her co-author Lisa Messeri, produced a paper describing the flaws in using AI systems to create supposed objectivity and increase productivity in scientific research. The conclusions map directly on to the unintelligent use of AI in newsrooms, which is becoming widespread – namely that we are in danger of “producing more and understanding less”.
Exactly the same types of limitations apply to the misuse of AI on news articles. “It reinforces a particular notion of objective truth,” says Crockett. “The notion that there is a capital T truth we can access in science is very brittle. It just provides an illusion of objectivity.” Mark Hansen, a statistician and director of the Brown Institute for Media Innovation at Columbia Journalism School (and therefore one of my colleagues) is equally sceptical at the idea of a bias detector. “In some ways it would be great if EVERYONE tried to build their own bias detector with AI, and really engage with the complexity of what that means,” he says. “Then they would find out how hard that is.”
For years now the new gatekeepers in the form of tech platforms have been funding research into the notion of “trust” in the media, at least partly as a way of distracting people from their own shortcomings. This research conveniently never touches on the decades-long effects of politicians and billionaires using social media to launch a relentless campaign against “trust” in the media, or what effects this type of standard propaganda campaign might have. Journalists such as the Nobel laureate Maria Ressa work in regimes that have effectively erased trust in independent media through disinformation campaigns. Ressa has been vocal about the dangers of “trust in media” research being weaponised. Now we have two of the most prominent owners of news organisations in America following a similar playbook. Businesspeople from other sectors might not understand the media well, or even the news organisations they own. But they do understand that it is bad business to trash your own currency in public.
This is not to say that there are not issues with the reliability, integrity and credibility of some or all of the news media most of the time. As an accountability mechanism it deserves as much scrutiny as the power structures it seeks to hold to account.
Rupert Murdoch, the blueprint for all media owners who aspire to influence governments, was fighting in court this week (so far unsuccessfully) to try to ensure his news assets remain ideologically aligned with the right, even from beyond the grave. At least there is a human in the loop here and a brutal honesty about Murdoch’s position. But the deployment of “trust” as a key failing in the US press and addressing “bias” as a way to fix it is either very naive or nakedly cynical. Moving a publication and its editorial line to the right is a simple process and this is what is happening with the Washington Post and LA Times. For the business interests of their owners to flourish they need to move to the right; Donald Trump is inherently hostile to organisations that challenge or report on him from anything other than a favourable position.
The first lesson in historian Timothy Snyder’s book On Tyranny is “Do not obey in advance”. In other words do not anticipate the desires of the powerful by making negative changes. The news ownership class in America is more unlikely to pay heed to this lesson now. And you can check that against your AI powered truth-o-meter.
